River hydrology and 1D–2D hydraulic modelling for flood projects
I estimate how big a design flood is, then model where it goes. At Hamstall Ridware the village gauge put the 1-in-100-year flow at 30.0 m³/s; pooling similar catchments gave 44.6 m³/s, half as much again. The hydrology was entirely my work. I built the 1D–2D hydraulic model jointly with an Environment Agency modelling colleague.
Real model output: the 2022 Flood Modeller 2D run at Hamstall Ridware, 6 m grid: a 1-in-100-year flood plus 29% for climate change spreads across the River Blithe and Trent floodplain, 149 ha at its peak. Replayed, not to time scale. How to read this: drag the slider to move through the flood.
Animated map: in the 2022 model of a 1% AEP flood with a 29% climate allowance, water spreads over 79 hours to cover 149 hectares of the River Blithe and Trent floodplain at Hamstall Ridware, Staffordshire.
- years of annual maximum flows in the gauge record I analysed for the River Blithe
- 50
- pooled 1-in-100-year flow to the village, against 30.0 m³/s from the gauge record alone
- 44.6m³/s
- grid of the linked 1D–2D model, built on 1 m LiDAR
- 5m
- buildings flooded at 1% AEP with the +29% climate allowance, up from 14 without it
- 21
About the author. Hisham Nuhu is a UK-based flood and disaster risk specialist. He worked at England’s Environment Agency from 2019 to 2025 and holds an MSc in Water Science, Policy and Management from the University of Oxford. He works remotely or in-country with governments, development banks, UN agencies and consulting firms, and in England with councils, communities, landowners and developers.
When does a project need a 1D–2D flood model rather than a global flood map?
When the decision turns on one bank, one bridge or one street. A 90 m global map is right for ranking Ghana’s 260 districts. It cannot say whether raising one bank keeps a village dry. At Hamstall the 2024 model could: the raised right-bank defence kept the village dry in every flood modelled, up to 0.1% AEP.
Global maps blur the edge of a flood, and the edge is where the buildings are. In my check against the 2020 White Volta flood in Ghana, moving the modelled edge by one 90 m cell changed the count of buildings inside the flood from 474 to 3,073.
So I use the two together. A national screen chooses the places; a 1D–2D model gives the depths a decision can rest on. At Hamstall the model also showed something no map could: opening the second arch of the road bridge gave no benefit upstream.
How do you estimate design flows when gauge records are short or unreliable?
Use several independent methods and let each do what it does best. The Hamstall gauge is bypassed in big floods, so its 50-year record understates them. I compared it with pooled statistics from similar catchments and the ReFH2 rainfall–runoff model, then scaled ReFH2 hydrograph shapes to the pooled peaks for each sub-catchment.
About three-quarters of the catchment drains through Blithfield Reservoir. I split the catchment at the dam and routed the upstream flood through the reservoir before it joined the smaller brooks. Then I added the Environment Agency’s climate allowances for the Trent valley: +15%, +17% and +29% on peak flow for the 2020s, 2050s and 2080s.
The tools are from the UK: the Flood Estimation Handbook, WINFAP pooling and ReFH2. The reasoning travels. In a basin with a short record I would pool it with similar gauged catchments in the region. Where a rating curve is doubtful at high flows, I would test it against flood marks and satellite flood extents. Where there is no gauge at all, a rainfall–runoff model driven by satellite rainfall or reanalysis, with its uncertainty stated, is the honest starting point.
The gauge record understates big floods; pooling similar catchments corrects it
How to read this: each dot is one year’s highest flow, placed by how rare it is. The shaded band is where water starts to bypass the gauge. Switch the pooled estimate on and off; hover, tap or use the arrow keys to read each year.
River Blithe at Hamstall Ridware, gauge 28002: 50 water years of annual maximum flow, a Gumbel fit to the gauge record, and the pooled estimate for the whole catchment.
Sources
Gauge data: National River Flow Archive station 28002 (UK Centre for Ecology & Hydrology; Environment Agency). Analysis: my flood estimation record (2022, updated 2025).
| Method | 1% AEP peak flow, m³/s | Against the pooled estimate |
|---|---|---|
| Gauge record alone (single-site statistics) | 30.0 | −33% |
| Pooled statistics from similar catchments (WINFAP) | 44.6 | – |
| Rainfall–runoff model (ReFH2) | 65.0 | +46% |
Show all 50 annual maximum flows as a table
| Water year | Annual maximum flow, m³/s |
|---|---|
| 1959/60 | 17.10 |
| 1960/61 | 17.30 |
| 1961/62 | 7.88 |
| 1962/63 | 2.74 |
| 1963/64 | 10.70 |
| 1968/69 | 17.30 |
| 1969/70 | 11.60 |
| 1970/71 | 23.70 |
| 1971/72 | 14.60 |
| 1972/73 | 7.70 |
| 1973/74 | 12.50 |
| 1974/75 | 8.56 |
| 1975/76 | 2.92 |
| 1976/77 | 20.10 |
| 1977/78 | 11.80 |
| 1978/79 | 16.80 |
| 1979/80 | 15.20 |
| 1980/81 | 15.40 |
| 1981/82 | 18.80 |
| 1982/83 | 15.40 |
| 1983/84 | 13.50 |
| 1984/85 | 9.72 |
| 1985/86 | 17.40 |
| 1994/95 | 19.80 |
| 1995/96 | 2.50 |
| 1996/97 | 9.07 |
| 1997/98 | 17.80 |
| 1998/99 | 15.50 |
| 1999/00 | 9.79 |
| 2000/01 | 22.00 |
| 2001/02 | 11.40 |
| 2002/03 | 17.40 |
| 2003/04 | 11.40 |
| 2004/05 | 9.49 |
| 2005/06 | 4.31 |
| 2006/07 | 16.10 |
| 2007/08 | 13.80 |
| 2008/09 | 13.20 |
| 2009/10 | 6.68 |
| 2010/11 | 3.76 |
| 2011/12 | 18.70 |
| 2012/13 | 18.10 |
| 2013/14 | 12.20 |
| 2014/15 | 4.34 |
| 2015/16 | 14.30 |
| 2016/17 | 10.20 |
| 2017/18 | 13.50 |
| 2018/19 | 9.30 |
| 2019/20 | 21.90 |
| 2020/21 | 19.70 |
How is a flood model calibrated and validated?
Against a flood that really happened, using whatever records exist. The Hamstall gauge cannot measure the biggest floods, so I checked the first model against residents’ own records of the February 2020 flood. Matching them meant raising channel roughness in wooded reaches from 0.03 to 0.075. For Ghana’s national map I used satellite images of a real flood.
A hindcast of a real event is the test a client finds most convincing, and checking the evidence this way finds weak links. Once I recognised, from the gauge’s own station notes, that floodwater bypasses it in big floods, I redid the river flows rather than defend the first estimate.
Where nobody recorded the flood, satellites can. Sentinel-1 radar sees water through cloud. Against its map of the September 2020 White Volta flood, the national flood map I used for Ghana scored a mean critical success index of 0.58 across five test areas.
What resolution and which software do you use?
Linked 1D–2D or full 2D models at 10–30 m for river reaches and towns, and finer where one bank decides the outcome: Hamstall ran at 5 m on 1 m LiDAR. The Hamstall models were in Flood Modeller and TUFLOW, in 1D and 2D. HEC-RAS 2D and SFINCS are other options for river and coastal 2D work.
- Engines
- Flood Modeller 1D linked to TUFLOW 2D (2024 model); Flood Modeller 2D on a 6 m grid (2022 model).
- Terrain and channel
- 5 m grid from 1 m LiDAR; a 20-cross-section channel survey I commissioned for the study.
- Boundaries
- The reservoir outflow, routed, plus the tributary brooks upstream; a fixed level on the River Trent downstream.
- Events
- 1% AEP with +15%, +17% and +29% climate allowances, and the 0.1% AEP flood.
- Analysis around the model
- Python, GDAL and QGIS for terrain, results and maps; R for statistics.
I build full 2D hydrodynamic models at 10–30 m resolution. I used TUFLOW and Flood Modeller 2D at Hamstall; HEC-RAS 2D and SFINCS are other options for river and coastal work.
Can you review a supplier’s model for a client or project team?
Yes. I check whether the design flows can be defended, whether the model was tested against a real flood, and whether its results support the decision being made. At the Environment Agency I reviewed developers’ flood risk assessments as a statutory consultee in London, and supported the delivery of two Flood Map for Planning models.
A review helps most before results are signed off: at inception, after the hydrology and before options are compared. Late reviews find the same problems, but by then they are expensive to fix.
How do model results feed warning and danger levels?
A model links a level at a gauge to what floods on the ground. That link sets the thresholds a warning service acts on. For ten incident weeks I worked the other end of that chain, as a Flood Warning Duty Officer checking forecasts against thresholds. At Hamstall a natural next step would be a local warning trigger from the modelled levels.
More on thresholds, return periods and impact-based forecasting: risk information for early warning.
How would you assess a river basin’s modelling and forecasting tools?
With four questions: what tools are used, whether they are fit for the decisions they support, what data and skills they depend on, and how they could be strengthened. At Hamstall the honest answers were that the gauge alone was not fit for design, and that three methods plus a real flood made the evidence fit.
| Question | At Hamstall Ridware |
|---|---|
| What tools were used? | The village gauge record; Flood Estimation Handbook statistics with WINFAP pooling and the ReFH2 rainfall–runoff model; routing through the reservoir; a linked Flood Modeller 1D and TUFLOW 2D model on 1 m LiDAR with a surveyed channel; Multi-Coloured Manual damages. |
| Were they fit for the decision? | Not on their own. The gauge is bypassed in floods, so single-site statistics came out a third below the pooled estimate. Reconciling three methods and checking the model against residents’ records made the evidence fit to choose between options. |
| What data and skills did they need? | LiDAR, a channel and structures survey, the reservoir operator’s data, residents’ records of a real flood, and a modeller able to document the chain for quality assurance. |
| How could they be strengthened? | Measure high flows at the gauge; keep model runs and routing workbooks together as a maintained model; link modelled levels to a local warning trigger. |
What do you hand over, and how do you train counterpart staff?
Everything someone else needs to rerun the work: the flood estimation record, the model files and run log, the survey and terrain references, and a report that explains every choice. Training is part of the hand-over. So far I have run two internal Environment Agency courses, for 10 and 3 colleagues, one using Hamstall as the worked example.
At Hamstall that meant a flood estimation record (2022, updated in 2025) and a technical report (December 2024), and a presentation of the results to the village.
What exactly is my record?
Said plainly, so you can match it to your terms of reference.
HamstallThe hydrology was entirely my work. I built the 1D–2D hydraulic model jointly with an Environment Agency modelling colleague.
Stafford CommonExtreme-flow hydrology for an urbanised catchment with several watercourses; no hydraulic model. The project.
2D on riversMy 2D modelling record is on rivers. Ghana’s coast is a static screening; 2D there is the next stage.
Other enginesHEC-RAS 2D and SFINCS are options I would use where a project calls for them; the models I have built, jointly at Hamstall, are in Flood Modeller and TUFLOW.
Questions people ask
Can you join a consulting firm’s bid as hydrologist or flood modeller?
Yes. Send me the terms of reference and the expert positions. I will say which ones my record fits and send a CV in the format the client asks for.
Can you model a river or a town in Ghana or West Africa?
Yes, as the detailed stage after a screening. Ghana’s national screen already points to places where a 2D model would change a decision. See flood risk work in Ghana and West Africa.
Do you build 2D coastal flood models?
I can build them at 10–30 m in TUFLOW or SFINCS-type models, but my completed 2D work is on rivers. For Ghana’s coast I built the screening stage. See coastal extreme water levels and inundation screening.
Can you work as an associate hydrologist for a UK consultancy?
Yes: FEH statistical and ReFH2 design flows, flows below reservoirs, and 1D–2D modelling. See FEH hydrology and modelling in England.
What does a 1% AEP flood mean?
A flood with a 1% chance of being equalled or exceeded in any year, often called the 1-in-100-year flood. It can happen in two years running.
Sources
- UK Centre for Ecology & Hydrology and Environment Agency. National River Flow Archive, station 28002: Blithe at Hamstall Ridware.
- UK Centre for Ecology & Hydrology. Flood Estimation Handbook.
- Wallingford HydroSolutions. ReFH 2 rainfall–runoff software.
- Environment Agency. Flood risk assessments: climate change allowances.
- TUFLOW and Flood Modeller: the 1D and 2D modelling software used at Hamstall Ridware.
- US Army Corps of Engineers, Hydrologic Engineering Center. HEC-RAS.
- Deltares. SFINCS.
- Copernicus Emergency Management Service. Global Flood Monitoring, Sentinel-1 flood extent, September 2020.
Work with me
Need design flows or a flood model you can defend?
Send the terms of reference or the decision the model must support. I will reply with how my record maps to the tasks, what data the work needs and when I am available, remotely or in-country.